RF Black Magic- Eliminating Fringing Capacitance

I'm not an RF guy. I don't think I've ever seen a VNA, and don't understand Smith charts. So this is just babbling:

If you clean slice off the end of a coax, the little shiny piece of the inner conductor is visible, and it has capacitance through the air to the shield and to the universe. But it doesn't have inductance, like the transmission line had. It looks like a lumped capacitor at the end of a transmission line.

Maybe cut away a little of the dielectric inside the coax, like a little cylinder or cone-shaped cut at the very end, to neutralize the external capacitance.

I am surprised that you see the streak at just 1 GHz. I guess VNAs must be pretty sensitive.

When you cut the coax, be careful to not squash it.

Looking at my N-calibration set for the HP8753D, the open is merely a hollow cylinder that screws onto the end connector, completed by a plug-in piece of plastic with a metal tip that mates with the centre conductor, so the central conductor ends much earlier than the screen, but the dielectric extends beyond it. I hope that's clear...

However, if you do a SOL single-port calibration, even with improvised short, open and load, all three should then end up as tight dots on the Smith chart. That says nothing about the quality of the standards.

Jeroen Belleman

Am 15.11.20 um 18:44 schrieb Cursitor Doom:

You may search the archives of the DG8SAQ vector network analyzer. They had long threads about cal kits.

The theme interests me less now , personally. I have a Z51 electronic cal kit for my R&S ZVB. I caught myself often before, that I ignored calibration, but VNA accuracy stands and falls with calibration.

Cheers, Gerhard

yeh point of the calibration is to make whatever you use as reference become it dots

A typical solution is to use a unity-gain buffer (emitter follower) to drive a shield around the node. Some variations (like autotransformer wound inductors to get around the below-unity voltage gain) can be a nuisance to install and trim, though. Controlling leakage is a useful side effect.

Inductive strays are harder, but usually a test jig just keeps connections short, and double-shielded coax isn't necessary.

Use the thinnest co-ax you can find. That way the fringe is smaller.

No, what you use as the reference should end up measuring as the values that the cal kit supplier said that the impedances of the standards are. In most cases they do not say that the standards are ideal (for the short and open at least), which is why the kits used to come with a floppy disk with the actual values on it.

Even if they could make an ideal open, there is usually an offset (length of transmission line) between the reference plane (where the impedance is measured) and the actual location of the open or short.

If the reference (cal kit) measured as a dot, and the manufacturer of the reference said that it should be something that is not a dot, then the calibration has been done wrongly.

My question relates to user calibration rather than instrument calibration. I don't think I've phrased the question very well at all on re-reading it so I'll take another shot at it during the week. Getting late here now so..... off to bed... ZZZzzzzzzzzzzz

We are not very gentle. You might mean gentile as in non-Jewish:

Elsewhere in this thread, you disclosed that the problem is with an open circuit termination that you have in your possession. Got any photos? Maker and model number? Is it built something like this? Open circuit: A coaxial open circuit standard has to use an enclosed approach to provide screening and avoid the pick-up of stray electromagnetic energy. At the open end of the inner conductor, a frequency-dependent fringing capacitance is formed. Even if an open standard could physically be constructed with zero length, the fringing capacitance would still result. This creates a negative imaginary part for S11 at higher frequencies as a result of the capacitive reactance.

I don't know of any method of building a termination that will solve the fringing capacitance problem over a wide bandwidth except to tune the load by adjusting the center conductor length so that it's resonant with the inherent fringing capacitance at the frequency range of interest. That will reduce the useful bandwidth of the VNA to a very narrow frequency range where it will only be accurate within that range.

Am 16.11.20 um 01:16 schrieb Cursitor Doom:

That thing is for user calibration. You just connect Port1, 2, 3, 4 ( I have only 1 & 2) to the box and the USB cable and the machine does the rest. No more swapping shorts, opens, 50R, throughs etc. PC3.5 / SMA has an official lifetime of 100 Plug-ins. and for an user calibration cycle you spend quite a few. Yes, we all stretch that. The ecal box makes all the necessary connections internally.

Gerhard

User calibration? I don't think I've ever had myself calibrated. Does it hurt?

The purpose of the calibration kit is to show the VNA what an open, a short and a load look like at the far end of this bit of cable used to connect to the DUT, so that it can remove the effects of that bit of cable. You usually want to know what your DUT looks like at *its* connector, not at the VNA connector.

Jeroen Belleman

Agreed, but my point is that most competent manufacturers of O/S/L cal kits are aware that the standards that they sell are not perfect, particularly the open, and they usually provide a model (or measurement) of the actual impedance at the reference plane. Where such a model is provided and is different from an ideal open, then after performing a calibration using that standard, and then re-connecting the "open" standard, one would expect the VNA to display the impedance that the standard is supposed to be, rather than an ideal open.

Yes, that makes sense. My N cal kit didn't come with such data.

Jeroen Belleman

First off, this calibration "error" is going to only effect phase readings on either S21 or S11. Is this several degrees of phase error something that is important to you? If not then move on.

Actually, I think it only effects S11 (or S22) readings. are your reflection measurements that critical?

May be another way to look at the problem : A VNA calibration kit is *not* a set of perfect open, shorts and loads, but a set of stable references, as close as possible to open/short/loads, with precisely measured characteristics. As far as I know, cal kit are always supplied with cal data (was on a floppy disk in the good old days... I guess they are now on an Eeprom inside the electronic cal kit).

Yours,

Robert Lacoste

Actually, it affects just about everything via the error correction. S-parameters, stability criteria ...

Google "12 term error correction".

regards, Gerhard

ings on either S21 or S11. Is this several degrees of phase error something that is important to you? If not then move on.

ction measurements that critical?

If these minor errors in the "open" calibration are going to make a differe nce in what he is doing then he is designing a very fragile circuit. Not t o say he is not designing a fragile circuit. But if you are not designing a fragile circuit then do not get wrapped about the axle over this.